Multiple alignment program for amino acid or nucleotide sequences

--add: Adding unaligned full-length sequence(s) into an existing alignment

% mafft --add new_sequences --reorder existing_alignment > output

• Sequences in new_sequences are ungapped and then aligned to existing_alignment. 
• new_sequences is a single multi-FASTA format file. 
• existing_alignment is a single multi-FASTA format file. 
• Gaps in existing_alignment are preserved, but the alignment length may be changed in the default setting (see example below).
• If the --keeplength option is given, then the alignment length is unchanged.  Insertions at the new sequences are deleted. 
• Add --mapout to see a correspondence table of positions, new_sequences.map, between before and after the calculation.  The --mapout option automatically turns on the --keeplength option, to keep the numbering of sites in the reference alignment (explanation added, 2016/Aug). 
• Omit --reorder to preserve the original sequence order. 
• Available in versions ≥6.811 (?); first described in Katoh & Frith 2012
• In versions ≥7.370, iterative refinement is performed if the --maxiterate n  flag is given, where n > 0.   The combination of iterative refinement and --keeplength is not supported (2017/Dec). New
• Accurate distance calculation (--localpair, --globalpair and --genafpair) can be applied to up to a few hundreds of sequences (2017/Dec). Updated
• Online version
• The difference between --addfull and --add will be explained later.

--addfragments: Adding unaligned fragmentary sequence(s) into an existing alignment

% mafft --auto --addfragments fragments --reorder --thread -1 existing_alignment > output

• Sequences in fragments are ungapped and then aligned to existing_alignment. 
• fragments is a single multi-FASTA format file. 
• existing_alignment is a single multi-FASTA format file. 
• Gaps in existing_alignment are preserved, but the alignment length may be changed in the default setting (see example below).
• If the --keeplength option is given, then the alignment length is unchanged.  Insertions at the fragmentary sequences are deleted. 
• Add --mapout to see a correspondence table of positions, fragments.map, between before and after the calculation.  The --mapout option automatically turns on the --keeplength option, to keep the numbering of sites in the reference alignment (explanation added, 2016/Aug). 
• --auto automatically switches algorithm according to data size.  Safer to always use this flag.  (added 2020/Sep)
• --multipair uses a high-cost (in time and memory usage) option.  Same as default.  Applicable to ∼<30,000 sites × ∼<1,000 sequences.
• --6merpair uses a low-cost option.
• Omit --reorder to preserve the original sequence order. 
• Described in Katoh & Frith 2012
• Can be used off-label to align closely-related sequences to a reference to build an MSA.
• Online version

Example

When existing_alignment is

>seq1
ACCDEFGHI-K
>seq2
A--DEFGHI-K

and a sequence to be added (newseq) is

>newseq
ACCDPQRSTEFG

then the result of   mafft --addfull (and mafft --addfragments, mafft --add) is

% mafft --addfull newseq existing_alignment

seq1            ACCD-----EFGHIK
seq2            A--D-----EFGHIK
newseq          ACCDPQRSTEFG---
                *  *     ***

The alignment length is changed (11→15) in this case, as PQRST is inserted (+5) and a gap-only column (between I and K) is removed (-1).

With the --keeplength option, the insertion PQRST is removed and the alignment length is kept unchanged (11→11).

% mafft --addfull newseq --keeplength existing_alignment

seq1            ACCDEFGHI-K
seq2            A--DEFGHI-K
newseq          ACCDEFG----
                *  ****

With the --mapout option, a correspondence table of positions is output to the newseq.map file.

% mafft --addfull newseq --mapout existing_alignment
% cat newseq.map

>newseq
# letter, position in the original sequence, position in the reference alignment
A, 1, 1
C, 2, 2
C, 3, 3
D, 4, 4
P, 5, -
Q, 6, -
R, 7, -
S, 8, -
T, 9, -
E, 10, 5
F, 11, 6
G, 12, 7

The --compactmapout option outputs the same information in a more compact form.  Available in versions 7.496 and higher. 2021/Dec

% mafft --addfull newseq --compactmapout existing_alignment
% cat newseq.map

# Insertions in the added sequences > Position in reference
>newseq
5P - 9T > 4v5

The --mapout and --compactmapout options automatically turn on the --keeplength option, to keep the consistency of the numbering of positions in the reference.

Bug information: Versions ≤7.154 had a bug in --addfragments.  When the sequences in the reference alignment were almost identical to each other, an incorrect result was occasionally returned due to this bug.  This bug has been fixed in version 7.157 (2014/Jun/10). Bug information: Versions 6.923 - 6.950 had a bug in the combination of --addfragments and --reorder.  The order of sequences in the output was incorrect.  This bug has been fixed in version 6.951 (2012/Oct/18). Bug information: Versions ≤6.815 had a problem in processing partial sequences.  When the new sequence has domains A and B but a part of sequences in the existing alignment lack domain B, domain B was sometimes not aligned.  This problem has been fixed in version 6.817 (2010/Aug/14).

Possible misalignment by versions ≤6.815:

existing alignment: AAAAAAAA---------BBBBBBBBB
existing alignment: AAAAAAAA---------BBBBBBBBB
existing alignment: AAAAAAAA------------------
new sequence:       AAAAAAAABBBBBBBBB---------

Fixed in versions ≥6.817:

existing alignment: AAAAAAAABBBBBBBBB
existing alignment: AAAAAAAABBBBBBBBB
existing alignment: AAAAAAAA---------
new sequence:       AAAAAAAABBBBBBBBB

Adding aligned sequences (profile) into an existing alignment

% mafft --addprofile aligned_sequences existing_alignment > output

• aligned_sequences are converted to a profile and then aligned to existing_alignment. 
• aligned_sequences should be a single multi-FASTA format file. 
• Both aligned_sequences and existing_alignment are preserved. 
• existing_alignment should be another multi-FASTA format file. 
• Accurate options (mafft-linsi, mafft-ginsi and mafft-einsi), instead of mafft, can be used for a few hundreds of sequences. 
• aligned_sequences must form a monophyletic cluster, as they are converted into a single profile.  Otherwise, the calculation stops.
• existing_alignment must form a paraphyletic cluster:
  

  or a monophyletic cluster:
  

Difference from the --seed option

• The --add and --addprofile options are much faster than the --seed option. 
• In the --seed option, the aligned letters in the given alignment are preserved but gaps are not necessarily preserved. 
• In the --add and --addfragments options, the given alignment (including the gap pattern) is completely preserved.

Difference from the mafft-profile program

The --addprofile option covers all the situations where the mafft-profile program was used.  Morever, the former is applicable to larger datasets than the latter.  Therefore, the mafft-profile program will be deleted in future releases. 

The mafft-profile program assumes that each profile separately forms a monophyletic cluster.